WO2022267381A1 - 一种呼吸机人机异步分类方法、系统、终端以及存储介质 - Google Patents

一种呼吸机人机异步分类方法、系统、终端以及存储介质 Download PDF

Info

Publication number
WO2022267381A1
WO2022267381A1 PCT/CN2021/137604 CN2021137604W WO2022267381A1 WO 2022267381 A1 WO2022267381 A1 WO 2022267381A1 CN 2021137604 W CN2021137604 W CN 2021137604W WO 2022267381 A1 WO2022267381 A1 WO 2022267381A1
Authority
WO
WIPO (PCT)
Prior art keywords
ventilator
data
breathing
man
respiratory
Prior art date
Application number
PCT/CN2021/137604
Other languages
English (en)
French (fr)
Inventor
谯小豪
李慧慧
熊富海
颜延
王磊
王博
Original Assignee
中国科学院深圳先进技术研究院
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 中国科学院深圳先进技术研究院 filed Critical 中国科学院深圳先进技术研究院
Publication of WO2022267381A1 publication Critical patent/WO2022267381A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/50—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for simulation or modelling of medical disorders
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/243—Classification techniques relating to the number of classes
    • G06F18/24323—Tree-organised classifiers
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients

Definitions

  • the invention belongs to the technical field of medical data processing, and in particular relates to a human-machine asynchronous classification method, system, terminal and storage medium of a ventilator.
  • Mechanical ventilation is the use of mechanical force to generate or enhance the patient's respiratory function with the help of a ventilator.
  • mechanical ventilation aims to improve oxygenation and reduce the patient's work of breathing and the load on the respiratory muscles to support the patient's improvement.
  • the inspiratory time and tidal volume of the ventilator must match the inspiratory time and tidal volume of the patient.
  • mechanical ventilation cannot adapt to different conditions of patients due to improper setting of ventilator parameters. Asynchronous (not synchronous) phenomena.
  • the clinically common human-machine asynchrony of the ventilator mainly occurs in the following three stages:
  • Phase 1 Inspiratory phase; major asynchronous events include trigger delay, inspiratory flow mismatch, short cycle, long cycle or reverse trigger, etc.
  • Phase II Transition phase from inhalation to exhalation; major asynchronous events include double triggering, respiratory muscle contractions, etc.
  • Phase III Expiratory phase; major asynchronous events include futile inspiratory effort, auto-triggering, respiratory muscle contraction, etc.
  • the more common of the above asynchronous events include invalid inspiratory effort, double triggering, and reverse triggering.
  • Ineffective inspiratory effort means that after the patient's breathing muscles are exerted, the trigger threshold of the ventilator is not reached, so that the ventilator does not produce air delivery, and the invalid trigger causes the patient's respiratory rate to be higher than the ventilator's frequency.
  • the double trigger is because the short inspiratory time of the ventilator is not coordinated with the patient's neural inspiratory time, so that the patient has two ventilator air delivery in one inspiratory effort, and the two air delivery volumes are very high, which can even make normal breathing Doubling the tidal volume, causing the patient's lungs to over-inflate, resulting in pulmonary barotrauma.
  • the automatic alarm system is used to provide early warning for the asynchronous event of the ventilator.
  • the existing automatic alarm system usually relies on simple thresholds, resulting in frequent false positives and causing great disturbance to medical staff.
  • the current data collection method for human-machine asynchronous research is to artificially remove noise interference after clinical collection. The data collection takes a long cycle and events, and human identification of interference factors is large, which will affect the accuracy of the data.
  • the present invention provides a human-machine asynchronous classification method, system, terminal and storage medium for a ventilator, aiming to solve one of the above-mentioned technical problems in the prior art at least to a certain extent.
  • the present invention provides the following technical solutions:
  • a ventilator man-machine asynchronous classification method comprising:
  • the breathing data after the labeling is input into the network model for training to obtain a trained man-machine asynchronous classification model
  • the human-machine asynchronous event of the ventilator is classified by the trained human-machine asynchronous classification model.
  • the technical solution adopted by the embodiment of the present invention also includes: the collection of breathing data under the man-machine asynchronous event simulated by the simulated lung and the ventilator includes:
  • the collected breathing data under human-machine asynchronous events include normal breathing, invalid inspiratory effort breathing and double-triggered breathing under three kinds of human-machine asynchronous events.
  • the technical solution adopted by the embodiment of the present invention also includes: the collection of respiratory data under the man-machine asynchronous event simulated by the simulated lung and ventilator also includes:
  • the collected respiratory data includes the respiratory data of the analog flow channel, tidal volume channel, airway pressure, alveolar pressure, pleural cavity pressure, heart pressure and bellows position channel.
  • the technical solution adopted by the embodiment of the present invention also includes: after the collection of the breathing data under the man-machine asynchronous event simulated by the simulated lung and the ventilator, it also includes:
  • the respiration data is preprocessed to obtain sample data corresponding to each man-machine asynchronous event.
  • the preprocessing of the breathing data includes:
  • the technical solution adopted in the embodiment of the present invention also includes: the feature extraction of the permutation deviation index on the respiratory data is specifically:
  • the eigenvalues of the displacement deviation index in the M-1 column respectively represent: the difference between the two channels of airflow and tidal volume, the difference between the two channels of tidal volume and alveolar pressure, and the difference between the two channels of alveolar pressure and airway pressure.
  • the technical solution adopted by the embodiment of the present invention further includes: the network model includes a decision tree or a random forest classifier.
  • a human-machine asynchronous classification system for a ventilator including:
  • Data collection module used to collect respiratory data under the asynchronous events of man-machine simulated by simulated lung and ventilator;
  • Feature extraction module used to perform permutation deviation index feature extraction on the respiratory data, and label the respiratory data according to the extracted permutation deviation index feature;
  • Asynchronous classification module used to input the labeled respiratory data into the network model for training, obtain a trained man-machine asynchronous classification model, and perform human-machine asynchronous events on the ventilator through the trained man-machine asynchronous classification model Classification.
  • a terminal includes a processor and a memory coupled to the processor, wherein,
  • the memory stores program instructions for implementing the ventilator-human-machine asynchronous classification method
  • the processor is configured to execute the program instructions stored in the memory to control the human-machine asynchronous classification of the ventilator.
  • Another technical solution adopted by the embodiment of the present invention is: a storage medium storing program instructions executable by a processor, and the program instructions are used to execute the human-machine asynchronous classification method for a ventilator.
  • the beneficial effect produced by the embodiment of the present invention is that the human-machine asynchronous classification method, system, terminal and storage medium of the embodiment of the present invention collect human-machine asynchronous events simulated by the simulated lung and the ventilator. After using the PDI feature to analyze the difference between the breathing data of adjacent channels, input the network model for training and output the human-computer asynchronous classification result.
  • the present invention has at least the following beneficial effects:
  • the collected respiratory data has less interference and is convenient to collect, which can be applied to the asynchronous classification of multiple cases.
  • Using PDI features to analyze the difference of breathing data of adjacent channels can well identify the difference of breathing data and further improve the accuracy of man-machine asynchronous classification.
  • Fig. 1 is the flow chart of the ventilator man-machine asynchronous classification method of the embodiment of the present invention
  • FIG. 2 is a schematic structural diagram of a human-machine asynchronous classification system for a ventilator according to an embodiment of the present invention
  • FIG. 3 is a schematic structural diagram of a terminal according to an embodiment of the present invention.
  • FIG. 4 is a schematic structural diagram of a storage medium according to an embodiment of the present invention.
  • FIG. 1 is a flow chart of a method for man-machine asynchronous classification of a ventilator according to an embodiment of the present invention.
  • the ventilator man-machine asynchronous classification method of the embodiment of the present invention comprises the following steps:
  • the simulated lung is the TestChest (intelligent cardiopulmonary bionic system) simulated lung
  • the ventilator is the Mindray SV300 ventilator
  • the simulated disease type is ARDS (Acute respiratory distress syndrome, acute respiratory distress syndrome) patients.
  • the TestChest simulated lung The frequency of the ventilator and the Mindray SV300 ventilator were set to 50HZ, the simulated respiratory rate was set to 21 times per minute, and the ventilation mode of the ventilator was CPAP/PSV mode. It can be understood that the present invention is also applicable to other types of simulated lungs, disease types and ventilators, and the parameters of the simulated lungs and ventilators can also be set according to actual applications.
  • the TestChest simulated lung can simulate the breathing of 15 channels, and the embodiment of the present invention only uses the simulated flow channel (Flow), tidal volume channel (Volume), airway pressure (Paw), alveolar pressure (Alveolar pressure), pleural cavity Intrapleural pressure, Cardiac pressure, and bellows position and other 7-channel respiratory data for man-machine asynchronous classification.
  • the embodiment of the present invention is also applicable to the man-machine asynchronous classification of respiratory data of other channels.
  • the collected multi-channel respiratory data includes the respiratory data of three human-machine asynchronous events: normal breathing, invalid inspiratory effort breathing, and double-triggered breathing. Taking the classification of the three asynchronous events of effortful breathing and double-trigger breathing as an example, it is also applicable to the classification of other human-machine asynchronous events such as automatic triggering and respiratory muscle contraction.
  • S2 Preprocessing the collected multi-channel respiratory data to obtain the sample data corresponding to each man-machine asynchronous event
  • preprocessing includes two parts: data segmentation and data labeling.
  • the data segmentation is specifically as follows: first, the peak and trough detection are performed on the respiratory data of the tidal volume channel, and the respiratory cycle in the respiratory data is obtained (each exhalation and inhalation is a respiratory cycle), and then, the respiratory data is processed according to the respiratory cycle. Segmentation processing to obtain segmented sample data; wherein, in order to ensure the balance of the sample data, the segmented sample data includes 150 normal breathing cycles, 150 invalid inspiratory effort breathing cycles, and 150 double-triggered breathing cycles. , the specific number of breathing cycles can be set according to the actual operation.
  • the data annotation is specifically: perform supplementary operation on the segmented sample data, set the sample data of each respiratory cycle to 98 data points, fill in zeros for the sample data that is less than 98 data points, and perform Each sample data is marked separately, and the sample data label of normal breathing is set to 1, the sample data label of double-triggered breathing is set to 2, and the sample data label of invalid inspiratory effort breathing is set to 3.
  • the specific number of data points and labeling methods can be set according to the actual operation.
  • PDI is a difference index, which is inversely proportional to the coupling strength between time series, and the present invention uses the PDI feature to identify the difference of respiratory data between channels.
  • the PDI feature extraction formula is as follows:
  • x and y respectively represent the time series of m-dimensional respiratory data of two adjacent channels in the sample data of normal respiration, double-trigger respiration or invalid inspiratory effort respiration, and the two time series x and y are mapped to vectors Xt and Yt respectively, where L represents the time interval in the permutation entropy of the two time series x and y, m represents the embedding dimension in the permutation entropy, and t represents time.
  • ⁇ i represents the symbol vector
  • n represents the number of occurrences of the time series x and y mapped to the specified sequence ⁇ i
  • N represents the total number of sampling points of the time series x and y
  • a high value of parameter ⁇ represents a super-Gaussian distribution
  • a low value represents a sub-Gaussian distribution
  • PDI(x,y) represents the PDI eigenvalue between vectors Xt and Yt.
  • the number of columns for PDI feature extraction and the number of feature values to be extracted can be set according to the number of channels for actually collecting respiratory data.
  • the network model includes a decision tree or random forest classifier.
  • the network model training process is as follows: 70% of the sample data of the three kinds of man-machine asynchronous events are used as the training set for model training, and 30% are used as the test set for model testing.
  • the accuracy rate, recall rate and F1 score of the human-machine asynchronous classification results output by the model are calculated through the decision tree or random forest algorithm to evaluate the model performance.
  • the respiratory data of the ARDS patient were collected for experimental verification, and the patient's ventilation cycle per minute was set to 21 times, and 150 normal breaths and 150 invalid inhalations of the patient were collected respectively.
  • the accuracy rate of the classification algorithm reaches 94.5%, the recall rate reaches 95.33%, and the F1 score is 0.949, and the accuracy rate of the random forest classification algorithm reaches 96.3%, the recall rate reaches 96.6%, and the F1 score is 0.963.
  • Experimental results show that the embodiment of the present invention can achieve higher classification accuracy.
  • the ventilator-human-machine asynchronous classification method of the embodiment of the present invention uses the simulated lung and the ventilator to simulate the human-machine asynchronous event, collects multi-channel respiratory data, and uses the PDI feature to analyze the difference between the respiratory data of adjacent channels, Input the network model for training and output human-machine asynchronous classification results.
  • the present invention has at least the following beneficial effects:
  • the collected respiratory data has less interference and is convenient to collect, which can be applied to the asynchronous classification of multiple cases.
  • Using PDI features to analyze the difference of breathing data of adjacent channels can well identify the difference of breathing data and further improve the accuracy of man-machine asynchronous classification.
  • FIG. 2 is a schematic structural diagram of a human-machine asynchronous classification system for a ventilator according to an embodiment of the present invention.
  • the ventilator man-machine asynchronous classification system 40 of the embodiment of the present invention includes:
  • Data collection module 41 used to collect breathing data under the man-machine asynchronous event simulated by the simulated lung and the ventilator;
  • Feature extraction module 42 used for performing PDI feature extraction on respiratory data, and labeling the respiratory data according to the extracted PDI feature;
  • Asynchronous classification module 43 used to input the labeled respiratory data into the network model for training, obtain a trained human-machine asynchronous classification model, and classify ventilator human-machine asynchronous events through the trained human-machine asynchronous classification model.
  • the man-machine asynchronous classification system for ventilator in the embodiment of the present invention collects multi-channel breathing data under the man-machine asynchronous event simulated by the simulated lung and the ventilator, uses the PDI feature to analyze the difference between the breathing data of adjacent channels, and then inputs The network model is trained and outputs human-machine asynchronous classification results.
  • the present invention has at least the following beneficial effects:
  • the collected respiratory data has less interference and is convenient to collect, which can be applied to the asynchronous classification of multiple cases.
  • Using PDI features to analyze the difference of breathing data of adjacent channels can well identify the difference of breathing data and further improve the accuracy of man-machine asynchronous classification.
  • FIG. 3 is a schematic diagram of a terminal structure according to an embodiment of the present invention.
  • the terminal 50 includes a processor 51 and a memory 52 coupled to the processor 51 .
  • the memory 52 stores program instructions for implementing the above-mentioned ventilator-human-machine asynchronous classification method.
  • the processor 51 is used to execute the program instructions stored in the memory 52 to control the human-machine asynchronous classification of the ventilator.
  • the processor 51 may also be referred to as a CPU (Central Processing Unit, central processing unit).
  • the processor 51 may be an integrated circuit chip with signal processing capabilities.
  • the processor 51 can also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components .
  • a general-purpose processor may be a microprocessor, or the processor may be any conventional processor, or the like.
  • FIG. 4 is a schematic structural diagram of a storage medium according to an embodiment of the present invention.
  • the storage medium in the embodiment of the present invention stores a program file 61 capable of realizing all the above-mentioned methods, wherein the program file 61 can be stored in the above-mentioned storage medium in the form of a software product, and includes several instructions to make a computer device (which can It is a personal computer, a server, or a network device, etc.) or a processor (processor) that executes all or part of the steps of the methods in various embodiments of the present invention.
  • a computer device which can It is a personal computer, a server, or a network device, etc.
  • processor processor
  • the aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disc, etc., which can store program codes. , or terminal devices such as computers, servers, mobile phones, and tablets.

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Public Health (AREA)
  • Medical Informatics (AREA)
  • Health & Medical Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • Primary Health Care (AREA)
  • General Health & Medical Sciences (AREA)
  • Epidemiology (AREA)
  • Pathology (AREA)
  • Databases & Information Systems (AREA)
  • Biomedical Technology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Evolutionary Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)

Abstract

一种呼吸机人机异步分类方法、系统、终端以及存储介质。所述方法包括：采集由模拟肺和呼吸机模拟的人机异步事件下的模拟流量通道、潮气量通道、气道压力、肺泡压、胸膜腔内压、心脏压力以及波纹管位置通道等多通道呼吸数据；对所述呼吸数据进行置换偏离指数(PDI)特征提取，并根据所提取的PDI特征为所述呼吸数据打上标签；将所述打标签后的呼吸数据输入网络模型进行训练，得到训练好的人机异步分类模型；通过所述训练好的人机异步分类模型对呼吸机人机异步事件进行分类。本方法采集的呼吸数据干扰较小，采集方便，并使用PDI特征进行相邻通道的呼吸数据的差异性分析，提高了人机异步分类的准确度。

Description

一种呼吸机人机异步分类方法、系统、终端以及存储介质 技术领域
本发明属医学数据处理技术领域，特别涉及一种呼吸机人机异步分类方法、系统、终端以及存储介质。
背景技术
机械通气是在呼吸机的帮助下，以机械力量产生或增强病人呼吸动作呼吸功能。在危重病人中，机械通气旨在改善氧合，减少病人呼吸工作和呼吸肌肉的负荷，以支持病人病情好转。在治疗过程中，为了满足病人的需求，呼吸机的吸气时间和潮气量必须和病人的吸气时间和潮气量相匹配。然而在实际应用中，由于呼吸机参数设置不当等原因，机械通气不能适应病人自身的不同病情，所以当病人的呼吸需求和呼吸机所提供的呼吸状态不匹配时，就会发生呼吸机人机异步(不同步)的现象。
临床上比较常见的呼吸机人机异步主要发生在以下三个阶段：
第一阶段：吸气期；主要异步事件包括触发延迟、吸气流量不匹配、短循环、长循环或反向触发等。
第二阶段：从吸气到呼气的过渡阶段；主要异步事件包括双触发、呼吸性肌肉收缩等。
第三阶段：呼气期；主要异步事件包括无效吸气努力、自动触发、呼吸肌肉收缩等。
以上异步事件中较为常见的包括无效吸气努力、双触发以及反向触发。无效吸气努力指的是病人呼吸肌肉用力后，未达到呼吸机的触发阈值，从而呼吸机未产生送气，无效的触发导致病人的呼吸频率高于呼吸机的频率。双触发是因为过短的呼吸机吸气时间与病人的神经吸气时间不协调，从而使得病人一次 吸气努力中出现两次呼吸机送气，而两次送气量非常高，甚至可以使正常呼吸的潮气量翻倍，从而造成病人的肺过度膨胀，导致肺气压伤。
目前，通过自动报警系统对呼吸机人机异步事件进行预警，然而现有的自动报警系统通常依赖于简单的阈值，导致频繁的假阳性，给医护人员造成很大的干扰。另外，目前关于人机异步的研究数据采集方式是通过临床采集后人为去除噪声干扰等，其数据采集耗费周期事件长，人为识别干扰因素大，会影响数据的准确性。
发明内容
本发明提供了一种呼吸机人机异步分类方法、系统、终端以及存储介质，旨在至少在一定程度上解决现有技术中的上述技术问题之一。
为了解决上述问题，本发明提供了如下技术方案：
一种呼吸机人机异步分类方法，包括：
采集由模拟肺和呼吸机模拟的人机异步事件下的呼吸数据；
对所述呼吸数据进行置换偏离指数特征提取，并根据所提取的置换偏离指数特征为所述呼吸数据打上标签；
将所述打标签后的呼吸数据输入网络模型进行训练，得到训练好的人机异步分类模型；
通过所述训练好的人机异步分类模型对呼吸机人机异步事件进行分类。
本发明实施例采取的技术方案还包括：所述采集由模拟肺和呼吸机模拟的人机异步事件下的呼吸数据包括：
所采集的人机异步事件下的呼吸数据包括正常呼吸、无效吸气努力呼吸和双触发呼吸三种人机异步事件下的呼吸数据。
本发明实施例采取的技术方案还包括：所述采集由模拟肺和呼吸机模拟的人机异步事件下的呼吸数据还包括：
所采集的呼吸数据包括模拟流量通道、潮气量通道、气道压力、肺泡压、 胸膜腔内压、心脏压力以及波纹管位置通道的呼吸数据。
本发明实施例采取的技术方案还包括：所述采集由模拟肺和呼吸机模拟的人机异步事件下的呼吸数据后还包括：
对所述呼吸数据进行预处理，分别得到各人机异步事件对应的样本数据。
本发明实施例采取的技术方案还包括：所述对所述呼吸数据进行预处理包括：
首先，对所述潮气量通道的呼吸数据进行波峰和波谷检测，获取呼吸数据中的呼吸周期，根据所述呼吸周期对呼吸数据进行分割处理，得到分割后的样本数据；所述分割的样本数据包括相同设定次数的正常呼吸周期、无效吸气努力呼吸周期以及双触发呼吸周期的呼吸数据；
然后，对所述分割后的样本数据进行补点操作，将每个呼吸周期的样本数据设为预设个数的数据点，对不够预设数据点个数的样本数据进行补零，并根据人机异步事件分别对每个样本数据进行标注。
本发明实施例采取的技术方案还包括：所述对所述呼吸数据进行置换偏离指数特征提取具体为：
分别对所述正常呼吸、无效吸气努力呼吸、双触发呼吸的样本数据的前M列进行置换偏离指数特征提取，分别得到所述正常呼吸、无效吸气努力呼吸、双触发呼吸的样本数据的M-1列置换偏离指数特征值；
所述M-1列置换偏离指数特征值分别代表：气流和潮气量两通道之间的差异性、潮气量和肺泡压两通道之间的差异性、肺泡压和气道压力两通道之间的差异性、气道压力和波纹管的位置两通道之间的差异性、波纹管的位置和胸膜腔内压两通道之间的差异性以及胸膜腔内压和心脏压力两通道之间的差异性。
本发明实施例采取的技术方案还包括：所述网络模型包括决策树或随机森林分类器。
本发明实施例采取的另一技术方案为：一种呼吸机人机异步分类系统，包括：
数据采集模块：用于采集由模拟肺和呼吸机模拟的人机异步事件下的呼吸数据；
特征提取模块：用于对所述呼吸数据进行置换偏离指数特征提取，并根据所提取的置换偏离指数征为所述呼吸数据打上标签；
异步分类模块：用于将所述打标签后的呼吸数据输入网络模型进行训练，得到训练好的人机异步分类模型，通过所述训练好的人机异步分类模型对呼吸机人机异步事件进行分类。
本发明实施例采取的又一技术方案为：一种终端，所述终端包括处理器、与所述处理器耦接的存储器，其中，
所述存储器存储有用于实现所述呼吸机人机异步分类方法的程序指令；
所述处理器用于执行所述存储器存储的所述程序指令以控制呼吸机人机异步分类。
本发明实施例采取的又一技术方案为：一种存储介质，存储有处理器可运行的程序指令，所述程序指令用于执行所述呼吸机人机异步分类方法。
相对于现有技术，本发明实施例产生的有益效果在于：本发明实施例的呼吸机人机异步分类方法、系统、终端以及存储介质通过采集由模拟肺和呼吸机模拟的人机异步事件下的多通道呼吸数据，采用PDI特征对相邻通道的呼吸数据进行差异性分析后，输入网络模型进行训练并输出人机异步分类结果。相对于现有技术，本发明至少具有以下有益效果：
1、采集的呼吸数据干扰较小，且采集方便，可适用于多病例的人机异步分类。
2、采集多通道呼吸数据进行分析，有利于提高人机异步分类准确度。
3、使用PDI特征进行相邻通道的呼吸数据的差异性分析，可以很好的识别出呼吸数据的差异性，进一步提高人机异步分类的准确度。
附图说明
图1是本发明实施例的呼吸机人机异步分类方法的流程图；
图2为本发明实施例的呼吸机人机异步分类系统结构示意图；
图3为本发明实施例的终端结构示意图；
图4为本发明实施例的存储介质的结构示意图。
具体实施方式
为了使本发明的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本发明进行进一步详细说明。应当理解，此处所描述的具体实施例仅用以解释本发明，并不用于限定本发明。
请参阅图1，是本发明实施例的呼吸机人机异步分类方法的流程图。本发明实施例的呼吸机人机异步分类方法包括以下步骤：
S1：采集由模拟肺和呼吸机模拟的人机异步事件下的多通道呼吸数据；
本步骤中，模拟肺为TestChest(智能心肺仿生系统)模拟肺，呼吸机为迈瑞SV300呼吸机，模拟的疾病类型为ARDS(Acute respiratory distress syndrome，急性呼吸窘迫综合症)病人，其中，TestChest模拟肺和迈瑞SV300呼吸机的频率分别设置为50HZ，模拟的呼吸频率设置为每分钟21次，呼吸机通气模式为CPAP/PSV模式。可以理解，本发明同样适用于其他类型的模拟肺、疾病类型以及呼吸机，模拟肺和呼吸机参数也可根据实际应用进行设置。
TestChest模拟肺可模拟15个通道的呼吸，本发明实施例仅采用其中的模拟流量通道(Flow)、潮气量通道(Volume)、气道压力(Paw)、肺泡压(Alveolar pressure)、胸膜腔内压(Intrapleural pressure)、心脏压力(Cardiac pressure)以及波纹管位置等7个通道的呼吸数据进行人机异步分类。本发明实施例同样适用于其他通道的呼吸数据的人机异步分类。
本发明实施例中，采集的多通道呼吸数据包括正常呼吸、无效吸气努力呼吸以及双触发呼吸三种人机异步事件的呼吸数据，可以理解，本发明仅以较为常见的正常呼吸、无效吸气努力呼吸以及双触发呼吸三种异步事件的分类为例， 同样适用于自动触发、呼吸肌肉收缩等其他人机异步事件的分类。
S2：对采集的多通道呼吸数据进行预处理，分别得到各人机异步事件对应的样本数据；
本步骤中，预处理包括数据分割和数据标注两个部分。数据分割具体为：首先，对潮气量通道的呼吸数据进行波峰和波谷检测，获取呼吸数据中的呼吸周期(每次呼气和吸气是一个呼吸周期)，然后，根据呼吸周期对呼吸数据进行分割处理，得到分割后的样本数据；其中，为了保证样本数据的平衡性，分割的样本数据分别包括150次正常呼吸周期、150次无效吸气努力呼吸周期以及150次双触发呼吸周期的呼吸数据，具体呼吸周期次数可根据实际操作进行设定。
数据标注具体为：对分割后的样本数据进行补点操作，将每个呼吸周期的样本数据设为98个数据点，对不够98个数据点的样本数据进行补零，并根据人机异步事件分别对每个样本数据进行标注，将正常呼吸的样本数据标签设置为1,双触发呼吸的样本数据标签设置为2,无效吸气努力呼吸的样本数据标签设置为3。具体数据点个数以及标注方式可根据实际操作进行设定。
S3：对样本数据进行置换偏离指数(Permutation Disalignment Index，PDI)特征提取，并根据提取的PDI特征为对应的样本数据打上标签；
本步骤中，PDI是一种差异性指标，与时间序列之间的耦合强度成反比，本发明利用PDI特征识别呼吸数据在各通道间的差异性。
具体的，PDI特征提取公式如下：
Xt＝[x(t),x(t+L),...,x(t+(m-1)L)]T    (1)
Yt＝[y(t),y(t+L),...,y(t+(m-1)L)]T    (2)
在公式(1)和(2)中，x、y分别代表正常呼吸、双触发呼吸或无效吸气努力呼吸的样本数据中两个相邻通道的m维呼吸数据时间序列，两个时间序列x和y分别映射到矢量Xt和Yt，L代表两个时间序列x和y的排列熵中的时间间隔，m代表排列熵中的嵌入维度，t代表时间。
p x,y(π i)＝n(π i)/(N-(m-1)L)   (3)
Figure PCTCN2021137604-appb-000001
在公式(3)中，πi代表符号矢量，n代表时间序列x和y映射到指定序列πi的出现个数，N代表时间序列x和y的采样点总数，概率p x,y(πi)代表时间序列x和y计算排列熵时出现相同排列的个数在总排列个数中所占的比例，符号矢量πi＝[τ1,τ2,...,τm]，τ代表时间延迟；
公式(4)中，参数α的高值代表超高斯分布，低值代表亚高斯分布，PDI(x,y)代表矢量Xt和Yt之间的PDI特征值。
进一步地，本发明实施例以分别对正常呼吸、无效吸气努力呼吸、双触发呼吸的样本数据的前M列进行PDI特征提取，分别得到三种人机异步事件的样本数据的M-1列PDI特征值；其中，本发明以M＝7为例，提取到的PDI特征值包括6列，第一列PDI特征值为气流和潮气量两通道之间的差异性，第二列PDI特征值为潮气量和肺泡压两通道之间的差异性，第三列PDI特征值为肺泡压和气道压力两通道之间的差异性，第四列PDI特征值为气道压力和波纹管的位置两通道之间的差异性，第五列PDI特征值为波纹管的位置和胸膜腔内压两通道之间的差异性，第六列PDI特征值为胸膜腔内压和心脏压力两通道之间的差异性。
可以理解，PDI特征提取的列数以及提取的特征值数量可根据实际采集呼吸数据的通道数进行设定。
S4：将打好标签的样本数据输入网络模型进行训练，得到训练好的人机异步分类模型；
本步骤中，网络模型包括决策树或随机森林分类器。网络模型训练过程具体为：分别将三种人机异步事件的样本数据中的70％作为用于模型训练的训练集，将30％作为用于模型测试的测试集。完成模型训练后，通过决策树或随机森林算法对模型输出的人机异步分类结果进行准确率、召回率及F1分数计算， 对模型性能进行评估。
S5：通过训练好的人机异步分类模型对呼吸机人机异步事件进行分类。
为了验证本发明实施例的可行性和有效性，采集了ARDS病人的呼吸数据进行实验验证，设定病人的每分钟通气周期为21次，分别采集病人的150次正常呼吸、150次无效吸气努力呼吸、150次双触发呼吸，对呼吸数据进行PDI特征提取后，通过人机异步分类模型输出人机异步分类结果，并分别利用决策树算法和随机森林算法对分类结果进行评估，其中决策树分类算法的准确率达到94.5％，召回率达到95.33％，F1得分为0.949，随机森林分类算法的准确率达到96.3％，召回率达到96.6％，F1得分为0.963。实验结果表明，本发明实施例可以达到较高的分类精度。
基于上述，本发明实施例的呼吸机人机异步分类方法通过采用模拟肺和呼吸机模拟人机异步事件，采集多通道呼吸数据，采用PDI特征对相邻通道的呼吸数据进行差异性分析后，输入网络模型进行训练并输出人机异步分类结果。相对于现有技术，本发明至少具有以下有益效果：
1、采集的呼吸数据干扰较小，且采集方便，可适用于多病例的人机异步分类。
2、采集多通道呼吸数据进行分析，有利于提高人机异步分类准确度。
3、使用PDI特征进行相邻通道的呼吸数据的差异性分析，可以很好的识别出呼吸数据的差异性，进一步提高人机异步分类的准确度。
请参阅图2，为本发明实施例的呼吸机人机异步分类系统结构示意图。本发明实施例的呼吸机人机异步分类系统40包括：
数据采集模块41：用于采集由模拟肺和呼吸机模拟的人机异步事件下的呼吸数据；
特征提取模块42：用于对呼吸数据进行PDI特征提取，并根据所提取的PDI特征为呼吸数据打上标签；
异步分类模块43：用于将打标签后的呼吸数据输入网络模型进行训练，得 到训练好的人机异步分类模型，通过训练好的人机异步分类模型对呼吸机人机异步事件进行分类。
本发明实施例的呼吸机人机异步分类系统通过采集由模拟肺和呼吸机模拟的人机异步事件下的多通道呼吸数据，采用PDI特征对相邻通道的呼吸数据进行差异性分析后，输入网络模型进行训练并输出人机异步分类结果。相对于现有技术，本发明至少具有以下有益效果：
1、采集的呼吸数据干扰较小，且采集方便，可适用于多病例的人机异步分类。
2、采集多通道呼吸数据进行分析，有利于提高人机异步分类准确度。
3、使用PDI特征进行相邻通道的呼吸数据的差异性分析，可以很好的识别出呼吸数据的差异性，进一步提高人机异步分类的准确度。
请参阅图3，为本发明实施例的终端结构示意图。该终端50包括处理器51、与处理器51耦接的存储器52。
存储器52存储有用于实现上述呼吸机人机异步分类方法的程序指令。
处理器51用于执行存储器52存储的程序指令以控制呼吸机人机异步分类。
其中，处理器51还可以称为CPU(Central Processing Unit，中央处理单元)。处理器51可能是一种集成电路芯片，具有信号的处理能力。处理器51还可以是通用处理器、数字信号处理器(DSP)、专用集成电路(ASIC)、现成可编程门阵列(FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。
请参阅图4，为本发明实施例的存储介质的结构示意图。本发明实施例的存储介质存储有能够实现上述所有方法的程序文件61，其中，该程序文件61可以以软件产品的形式存储在上述存储介质中，包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)或处理器(processor)执行本发明各个实施方式方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、只读存储器(ROM，Read-Only Memory)、随机存取存储器(RAM， Random Access Memory)、磁碟或者光盘等各种可以存储程序代码的介质，或者是计算机、服务器、手机、平板等终端设备。
对所公开的实施例的上述说明，使本领域专业技术人员能够实现或使用本发明。对这些实施例的多种修改对本领域的专业技术人员来说将是显而易见的，本发明中所定义的一般原理可以在不脱离本发明的精神或范围的情况下，在其它实施例中实现。因此，本发明将不会被限制于本发明所示的这些实施例，而是要符合与本发明所公开的原理和新颖特点相一致的最宽的范围。

Claims (10)

  1. 一种呼吸机人机异步分类方法，其特征在于，包括：
    采集由模拟肺和呼吸机模拟的人机异步事件下的呼吸数据；
    对所述呼吸数据进行置换偏离指数特征提取，并根据所提取的置换偏离指数特征为所述呼吸数据打上标签；
    将所述打标签后的呼吸数据输入网络模型进行训练，得到训练好的人机异步分类模型；
    通过所述训练好的人机异步分类模型对呼吸机人机异步事件进行分类。
  2. 根据权利要求1所述的呼吸机人机异步分类方法，其特征在于，所述采集由模拟肺和呼吸机模拟的人机异步事件下的呼吸数据包括：
    所采集的人机异步事件下的呼吸数据包括正常呼吸、无效吸气努力呼吸和双触发呼吸三种人机异步事件下的呼吸数据。
  3. 根据权利要求2所述的呼吸机人机异步分类方法，其特征在于，所述采集由模拟肺和呼吸机模拟的人机异步事件下的呼吸数据还包括：
    所采集的呼吸数据包括模拟流量通道、潮气量通道、气道压力、肺泡压、胸膜腔内压、心脏压力以及波纹管位置通道的呼吸数据。
  4. 根据权利要求3所述的呼吸机人机异步分类方法，其特征在于，所述采集由模拟肺和呼吸机模拟的人机异步事件下的呼吸数据后还包括：
    对所述呼吸数据进行预处理，分别得到各人机异步事件对应的样本数据。
  5. 根据权利要求4所述的呼吸机人机异步分类方法，其特征在于，所述对所述呼吸数据进行预处理包括：
    首先，对所述潮气量通道的呼吸数据进行波峰和波谷检测，获取呼吸数据中的呼吸周期，根据所述呼吸周期对呼吸数据进行分割处理，得到分割后的样本数据；所述分割的样本数据包括相同设定次数的正常呼吸周期、无效吸气努力呼吸周期以及双触发呼吸周期的呼吸数据；
    然后，对所述分割后的样本数据进行补点操作，将每个呼吸周期的样本数据设为预设个数的数据点，对不够预设数据点个数的样本数据进行补零，并根据人机异步事件分别对每个样本数据进行标注。
  6. 根据权利要求4所述的呼吸机人机异步分类方法，其特征在于，所述对所述呼吸数据进行置换偏离指数特征提取具体为：
    分别对所述正常呼吸、无效吸气努力呼吸、双触发呼吸的样本数据的前M列进行置换偏离指数特征提取，分别得到所述正常呼吸、无效吸气努力呼吸、双触发呼吸的样本数据的M-1列置换偏离指数特征值；
    所述M-1列置换偏离指数特征值分别代表：气流和潮气量两通道之间的差异性、潮气量和肺泡压两通道之间的差异性、肺泡压和气道压力两通道之间的差异性、气道压力和波纹管的位置两通道之间的差异性、波纹管的位置和胸膜腔内压两通道之间的差异性以及胸膜腔内压和心脏压力两通道之间的差异性。
  7. 根据权利要求1至6任一项所述的呼吸机人机异步分类方法，其特征在于，所述网络模型包括决策树或随机森林分类器。
  8. 一种呼吸机人机异步分类系统，其特征在于，包括：
    数据采集模块：用于采集由模拟肺和呼吸机模拟的人机异步事件下的呼吸数据；
    特征提取模块：对所述呼吸数据进行置换偏离指数特征提取，并根据所提取的置换偏离指数特征为所述呼吸数据打上标签；
    异步分类模块：用于将所述打标签后的呼吸数据输入网络模型进行训练，得到训练好的人机异步分类模型，通过所述训练好的人机异步分类模型对呼吸机人机异步事件进行分类。
  9. 一种终端，其特征在于，所述终端包括处理器、与所述处理器耦接的存储器，其中，
    所述存储器存储有用于实现权利要求1-7任一项所述的呼吸机人机异步分类方法的程序指令；
    所述处理器用于执行所述存储器存储的所述程序指令以控制呼吸机人机异步分类。
  10. 一种存储介质，其特征在于，存储有处理器可运行的程序指令，所述程序指令用于执行权利要求1至7任一项所述呼吸机人机异步分类方法。
PCT/CN2021/137604 2021-06-25 2021-12-13 一种呼吸机人机异步分类方法、系统、终端以及存储介质 WO2022267381A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202110708905.4 2021-06-25
CN202110708905.4A CN113539501A (zh) 2021-06-25 2021-06-25 一种呼吸机人机异步分类方法、系统、终端以及存储介质

Publications (1)

Publication Number Publication Date
WO2022267381A1 true WO2022267381A1 (zh) 2022-12-29

Family

ID=78096697

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/137604 WO2022267381A1 (zh) 2021-06-25 2021-12-13 一种呼吸机人机异步分类方法、系统、终端以及存储介质

Country Status (2)

Country Link
CN (1) CN113539501A (zh)
WO (1) WO2022267381A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115814222A (zh) * 2023-01-17 2023-03-21 中国科学院深圳先进技术研究院 混合机械通气模式下人机不同步波形识别方法及相关设备

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113539501A (zh) * 2021-06-25 2021-10-22 中国科学院深圳先进技术研究院 一种呼吸机人机异步分类方法、系统、终端以及存储介质
CN113539398A (zh) * 2021-06-25 2021-10-22 中国科学院深圳先进技术研究院 一种呼吸机人机异步分类方法、系统、终端以及存储介质
CN113951868B (zh) * 2021-10-29 2024-04-09 北京富通东方科技有限公司 机械通气患者人机不同步检测方法及装置
CN114191665A (zh) * 2021-12-01 2022-03-18 中国科学院深圳先进技术研究院 机械通气过程中人机异步现象的分类方法和分类装置
CN114288500B (zh) * 2021-12-02 2024-03-19 中国科学院深圳先进技术研究院 一种基于模糊熵特征提取的人机通气异步检测模型及装置
CN114216712B (zh) * 2021-12-15 2024-03-08 深圳先进技术研究院 一种机械通气人机异步数据获取方法、检测方法及其设备

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109893732A (zh) * 2019-02-28 2019-06-18 杭州智瑞思科技有限公司 一种基于循环神经网络的机械通气人机不同步检测方法
CN111383764A (zh) * 2020-02-25 2020-07-07 山东师范大学 一种机械通气驱动压与呼吸机相关事件的相关性检测系统
CN111563451A (zh) * 2020-05-06 2020-08-21 浙江工业大学 基于多尺度小波特征的机械通气无效吸气努力识别方法
WO2021122011A1 (en) * 2019-12-16 2021-06-24 Consorci Corporació Sanitària Parc A device and method for respiratory monitoring in mechanically ventilated patients
CN113539501A (zh) * 2021-06-25 2021-10-22 中国科学院深圳先进技术研究院 一种呼吸机人机异步分类方法、系统、终端以及存储介质

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP3706845B1 (en) * 2017-11-09 2023-10-18 Autonomous Healthcare, Inc. Clinical decision support system for patient-ventilator asynchrony detection and management
DE102020004319A1 (de) * 2019-07-26 2021-01-28 Löwenstein Medical Technology S.A. Atemzuganalysator, Beatmungsgerät und Verfahren zur Atemzuganalyse
CN111738305B (zh) * 2020-05-29 2022-06-24 浙江大学 一种基于dba-dtw-knn的机械通气人机不同步快速识别方法
CN112932506A (zh) * 2021-02-01 2021-06-11 中国科学院深圳先进技术研究院 一种推拿疗效评估方法、装置、系统、以及存储介质

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109893732A (zh) * 2019-02-28 2019-06-18 杭州智瑞思科技有限公司 一种基于循环神经网络的机械通气人机不同步检测方法
WO2021122011A1 (en) * 2019-12-16 2021-06-24 Consorci Corporació Sanitària Parc A device and method for respiratory monitoring in mechanically ventilated patients
CN111383764A (zh) * 2020-02-25 2020-07-07 山东师范大学 一种机械通气驱动压与呼吸机相关事件的相关性检测系统
CN111563451A (zh) * 2020-05-06 2020-08-21 浙江工业大学 基于多尺度小波特征的机械通气无效吸气努力识别方法
CN113539501A (zh) * 2021-06-25 2021-10-22 中国科学院深圳先进技术研究院 一种呼吸机人机异步分类方法、系统、终端以及存储介质

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115814222A (zh) * 2023-01-17 2023-03-21 中国科学院深圳先进技术研究院 混合机械通气模式下人机不同步波形识别方法及相关设备
CN115814222B (zh) * 2023-01-17 2023-04-14 中国科学院深圳先进技术研究院 混合机械通气模式下人机不同步波形识别方法及相关设备

Also Published As

Publication number Publication date
CN113539501A (zh) 2021-10-22

Similar Documents

Publication Publication Date Title
WO2022267381A1 (zh) 一种呼吸机人机异步分类方法、系统、终端以及存储介质
WO2022267382A1 (zh) 一种呼吸机人机异步分类方法、系统、终端以及存储介质
CN109893732B (zh) 一种基于循环神经网络的机械通气人机不同步检测方法
Hsu et al. Benchmarking of eight recurrent neural network variants for breath phase and adventitious sound detection on a self-developed open-access lung sound database—HF_Lung_V1
KR102267105B1 (ko) 딥러닝 기반의 수면다원 검사장치 및 그 방법
WO2022242123A1 (zh) 机械通气人机异步检测方法、装置及计算机可读存储介质
CN110348335A (zh) 行为识别的方法、装置、终端设备及存储介质
CN110801221A (zh) 基于无监督特征学习的睡眠呼吸暂停片段检测方法及设备
CN113642512B (zh) 呼吸机人机异步检测方法、装置、设备及存储介质
CN108836269A (zh) 一种融合心率呼吸体动的自动睡眠分期方法
WO2019113222A1 (en) A data processing system for classifying keyed data representing inhaler device operation
WO2022006150A1 (en) Systems and methods for respiratory support recommendations
CN113941061A (zh) 一种人机不同步识别方法、系统、终端以及存储介质
WO2023097780A1 (zh) 机械通气过程中人机异步现象的分类方法和分类装置
Pham et al. Respiratory artefact removal in forced oscillation measurements: a machine learning approach
WO2024152392A1 (zh) 混合机械通气模式下人机不同步波形识别方法及相关设备
Abushakra et al. Efficient frequency-based classification of respiratory movements
CN116469568A (zh) 基于呼吸训练采集数据的呼吸训练模型创建方式及设备
CN109589117A (zh) 一种基于大数据的呼吸内科用监护仪控制方法及系统
WO2023060478A1 (zh) 一种人机不同步识别方法、系统、终端以及存储介质
CN115414026A (zh) 一种基于流速波形的呼吸自动切分方法及其系统
Avci et al. Sleep apnea detection using adaptive neuro fuzzy inference system
CN113730755B (zh) 一种基于注意力机制的机械通气人机异步检测识别方法
WO2023097785A1 (zh) 一种基于模糊熵特征提取的人机通气异步检测模型及装置
Fahed et al. Random forest classification of breathing phases from audio signals recorded using mobile devices

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21946855

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21946855

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 21946855

Country of ref document: EP

Kind code of ref document: A1